Graphduty: monitoring via graphite

Abstract: At knewton, we've managed to get pervasive system statistics into graphite across all of our deployments. In-house software, system stats, and servers (cassandra, rabbitmq, etc.) all report into graphite.

We've leveraged this by writing a simple tool, called graphduty, which will read configured sets of stats from graphite, and edge-trigger on events and notify pagerduty so that they can be addressed without having to create monster dashboards.

Speaker: Peter Norton